[WIP]
Pytorch implementation of Performer self attention module from the paper "Rethinking Attention with Performers".
This implementation is based on available code in https://github.com/google-research/google-research/tree/master/performer/fast_self_attention
Run tests to compare against Google's implementation in JAX.
python test.py
Requires JAX to be installed.